Elsevier

Applied Soft Computing

Volume 13, Issue 7, July 2013, Pages 3225-3233
Applied Soft Computing

Short-term wind speed forecasting based on a hybrid model

https://doi.org/10.1016/j.asoc.2013.02.016Get rights and content

Abstract

Wind power is currently one of the types of renewable energy with a large generation capacity. However, operation of wind power generation is very challenging because of the intermittent and stochastic nature of the wind speed. Wind speed forecasting is a very important part of wind parks management and the integration of wind power into electricity grids. As an artificial intelligence algorithm, radial basis function neural network (RBFNN) has been successfully applied into solving forecasting problems. In this paper, a novel approach named WTT–SAM–RBFNN for short-term wind speed forecasting is proposed by applying wavelet transform technique (WTT) into hybrid model which hybrids the seasonal adjustment method (SAM) and the RBFNN. Real data sets of wind speed in Northwest China are used to evaluate the forecasting accuracy of the proposed approach. To avoid the randomness caused by the RBFNN model or the RBFNN part of the hybrid model, all simulations in this study are repeated 30 times to get the average. Numerical results show that the WTT–SAM–RBFNN outperforms the persistence method (PM), multilayer perceptron neural network (MLP), RBFNN, hybrid SAM and RBFNN (SAM–RBFNN), and hybrid WTT and RBFNN (WTT–RBFNN). It is concluded that the proposed approach is an effective way to improve the prediction accuracy.

Highlights

► This paper proposes a novel hybrid approach for wind speed forecasting. ► WTT is applied to remove the useless information in wind speed series. ► SAM and RBFNN are applied to predict the seasonal component and trend component in the wind speed series respectively. ► The real datasets in Northwest China are used to demonstrate the forecasting accuracy of the proposed approach. ► The empirical results show that the proposed approach can be an effective way and very promising.

Introduction

In recent years, due to global environmental pollution issues, renewable energy (such as wind, solar, geothermal, biomass, tidal, and hydropower) has received increasing attention. Wind power is one of the cleanest renewable energy sources that produce no greenhouse gases, has no effect on climate change, and produces little environmental impacts, and the energy generated from the wind has been well recognized as environmentally friendly, socially beneficial, and economically competitive for many applications [1]. However, the power generation from the wind has been plagued by the intermittent and stochastic nature of wind source, and thus it is still a less reliable source and difficult to be integrated into power grid systems [2]. This problem can be significantly mitigated if the operation of wind farm can be controlled based on the accurate information of dynamic wind speed forecasting. Accurate short-term wind speed forecasting can reduce voltage and frequency fluctuations due to variation in wind power and unacceptable shocks in the conventional power units caused by a sudden cut-off of wind power resulting from excessive wind speeds [3], [4]. As a result, accurate forecasting of the short-term wind speed is a critical issue for improving both the reliability of a wind power generation system and the integration of wind energy into the power system [5].

Short-term wind speed forecasting can be made in the order of several days and also from minutes to hours [6]. Usually, hourly forecasts of expected winds are helpful in dispatching decision-making, daily forecasts of hourly winds are useful for the load scheduling strategy, and weekly forecasts of day-to-day winds greatly facilitate maintenance scheduling [2], [7]. Many methods on the short-term wind speed forecasting have been developed over the past two decades. Generally, these methods used in the literature can be divided into two categories: statistical models and artificial intelligence models (AI). Statistical models are identical to the direct random time-series model, including auto regressive (AR), and auto regressive integrated moving average (ARIMA) models [9]. They do well in short-term forecasts, and have been tested in short-term wind speed forecasts [3], [8], [10], [11], [12], [13], [14]. However, they are not perfect in forecasting. First, most of statistical models assume that the wind speed data is normally distributed, but it is well known that wind speed series is not a normally distributed [3]. Second, the intermittent and stochastic characteristic of wind speed series need more complex functions for capturing the nonlinear relations, but these models are based on the assumption that a linear correlation structure exists among time series values [15]. As a result, the wind speed series is difficult to be predicted accurately. To overcome this limitation of statistical approaches, the AI models, mainly including artificial neural networks (ANNS), have attracted more attention for accurate short-term wind speed forecasting and have also been determined to be more accurate as compared to traditional statistical models [4], [8], [16], [17], [18], [19], [11], [20], [21], [22], [23], [24], [25].

ANNs are data-driven and non-parametric models, and can be useful techniques for wind speed forecasting due to their ability to capture subtle functional relationships among the empirical data even though the underlying relationships are unknown or hard to describe. As a special neural network, radial basis function neural networks (RBFNNs) have been extensively studied by researchers in nonlinear identification and time series forecasting areas [26], [27], [28], [29]. They have the ability to rapidly learn complex patterns and tendencies presented in data, with fast adaptation to changes. It has been proved that a RBFNN can approximate arbitrarily well any multivariate continuous function on a compact domain if a sufficient number of radial basis function units are given [30], [31]. It motivates this study of using RBFNN for the short-term wind speed forecasting in this paper.

Wind speed series are non-stationary and highly-noisy due to wind being a weather driven renewable resource which mainly depends on the climate system. Forecasting the wind speed data with the noisy directly is usually subject to large errors. The wavelet transform technique (WTT) is an essential tool for data pre-processing and has been widely used in de-noising and extracting the basic characteristics from the non-stationary time series [32]. For this reason, this WTT is applied to the wind speed series for de-noising in the paper. On the other hand, the seasonal and trend variations are the two most commonly encountered phenomena in the wind speed series. The seasonal component information is often neglected in most wind speed series forecasting, whose fluctuation causes large deviation to the forecasting. According to Zhang and Qi [33], de-seasonalization can dramatically reduce forecasting errors in many seasonal time series. The seasonal adjustment method (SAM) which can separate the seasonal component and trend component from the time series can help estimate the trend and make short-term forecasting more efficient, so the SAM is applied for the preprocessing of the wind speed data in this paper.

The WTT can effectively remove the useless information in wind speed series. The SAM is mainly applied to forecast the seasonal component and the RBFNN is adopted to forecast the trend component in the wind speed datasets. Considering the actual features of wind speed series, in this paper, a novel approach named WTT–SAM–RBFNN for short-term wind speed forecasting is proposed by applying WTT into hybrid model which hybrids the SAM and RBFNN. In such a model, the raw wind speed data are first decomposed into an approximate part associated with low frequencies and a detail part associated with high frequencies by the WTT. Second, the SAM separates the seasonal component and trend component from the low-frequency signal series, and simulates and predicts the seasonal component. Third, the trend component is predicted by the RBFNN, and the forecasting value of the raw wind speed data is got by multiplying the seasonal index to the forecasting value of the trend component. Finally, the mean hourly wind speed data about one month in Wuwei and Minqin in China are used as illustrative examples to evaluate the forecasting accuracy of the proposed approach. The empirical results also clearly show that the WTT–SAM–RBFNN outperforms the persistence method (PM), multilayer perceptron neural network (MLP), RBFNN, hybrid SAM and RBFNN (SAM–RBFNN), and hybrid WTT and RBFNN (WTT–RBFNN). It is concluded that the proposed approach is an effective way to improve the forecasting accuracy.

The rest of this paper is organized as follows. Section 2 presents the WTT–SAM–RBFNN approach for forecasting the wind speed. Section 3 provides the evaluation criteria and presents the numerical results from two real world cases study. Finally, Section 4 outlines the conclusions.

Section snippets

Proposed approach

The proposed WTT–SAM–RBFNN approach, which applies the WTT into a hybrid model which hybrids SAM and RBFNN, is proposed for forecasting short-term wind speed in this paper. The algorithm is described as follows and the flowchart is shown in Fig. 1. The methods used in the WTT–SAM–RBFNN approach are briefly introduced in the following sections.

  • Step 1: Wavelet de-noising. The raw wind speed data series are decomposed into a low-frequency component and a high-frequency component by the WTT. The

Data sets

The Wuwei and Minqin region, which are located in the Gansu Province in China, has abundant wind resources due to its geographical characteristics. The two regions have the potential to be a valuable wind farm site. To investigate actual wind power potential, it is highly desirable to forecast wind speeds in the two regions.

In this paper, the 24 hourly mean data for about one month in Wuwei and Minqin are selected as illustrative examples to evaluate the proposed approach. Fig. 3 shows the

Conclusions

Taken the highly-noisy and seasonal characteristics of the wind speed series into consideration, a hybrid approach combining WTT, SAM with RBFNN is proposed for wind speed forecasting in this paper. The basic idea of the hybrid model is to remove the noise and consider the seasonal factors in raw data series. As an example, the mean hourly wind speed data about one month in Wuwei and Minqin are applied to validate the forecasting performance of the proposed approach. The results show that the

Acknowledgments

This research was supported by the National Basic Research Program of China ‘973’ Program (Grant No. 2012CB956200), the National Natural Science Foundation of China (Grant No. 71171102), the Opening Fund of Key Laboratory of Land Surface Process and Climate Change in Cold and Arid Regions, CAS, and the Scholarship Award for Excellent Doctoral granted by Lanzhou University.

References (53)

  • M.A. Mohandes et al.

    Support vector machines for wind speed prediction

    Renew Energy

    (2004)
  • P. Flores et al.

    Application of a control algorithm for wind speed prediction and active power generation

    Renew Energy

    (2005)
  • M. Carolin Mabel et al.

    Analysis of wind power generation and prediction using ANN: a case study

    Renew Energy

    (2008)
  • S. Salcedo-Sanz et al.

    Accurate short-term wind speed prediction by exploiting diversity in input data using banks of artificial neural networks

    Neurocomputing

    (2009)
  • N. Qu et al.

    Application of GA-RBF networks to the nondestructive determination of active component in pharmaceutical powder by NIR spectroscopy

    Journal of the Taiwan Institute of Chemical Engineers

    (2009)
  • V.M. Rivas et al.

    Evolving RBF neural networks for time-series forecasting with EvRBF

    Information Sciences

    (2004)
  • W. Shen et al.

    Forecasting stock indices using radial basis function neural networks optimized by artificial fish swarm algorithm

    Knowledge-Based Systems

    (2011)
  • C.L. Lin et al.

    Improving the generalization performance of RBF neural networks using a linear regression technique

    Expert Systems with Applications

    (2009)
  • B. Cannas et al.

    Data preprocessing for river flow forecasting using neural networks: Wavelet transforms and data partitioning

    Physics and Chemistry of the Earth

    (2006)
  • G.P. Zhang et al.

    Neural network forecasting for seasonal and trend time series

    European Journal of Operational Research

    (2005)
  • S.T. Li et al.

    Knowledge discovery in financial investment for forecasting and trading strategy through wavelet-based SOM networks

    Expert Systems with Applications

    (2008)
  • J.Z. Wang et al.

    Forecasting stock indices with back propagation neural network

    Expert Systems with Applications

    (2011)
  • C.J. Lu

    Integrating independent component analysis-based denoising scheme with neural network for stock price prediction

    Expert Systems with Applications

    (2010)
  • Z.H. Guo et al.

    A case study on a hybrid wind speed forecasting method using BP neural network

    Knowledge-Based Systems

    (2011)
  • C.C. Lee et al.

    Noisy time series prediction using M-estimator based robust radial basis function neural networks with growing and pruning techniques

    Expert Systems with Applications

    (2009)
  • M.S. Miranda et al.

    One-hour-ahead wind speed prediction using a bayesian methodology

  • Cited by (0)

    View full text